Steam Flow Reads Half of the Heat Balance at Part Load
A 12-inch steam header feeds a reboiler. Flow is measured with an orifice plate and a DP transmitter ranged 0–150 inH₂O, which the orifice datasheet says corresponds to 45,000 lb/h at design conditions (250 psig, 500 °F). The 4–20 mA signal lands on a DCS analog input; the HMI point is scaled 0–45,000 lb/h and drives both the indication and the totalizer. The transmitter was calibrated 0–150 inH₂O at the bench. Three months after commissioning, the plant heat balance says the reboiler receives about twice the steam that the flow indication shows when the boiler is at 50% of design firing; at design firing the indication and the heat balance agree within 2%. There is no pressure or temperature compensation anywhere in the loop. The header pressure controller typically holds 250 psig ±8 psi.
1. Diagnose the error, showing the arithmetic that reproduces the observation, and explain why the fault survived commissioning.
Read the worked answer
Diagnosis: the square-root extraction is missing. DP across an orifice is proportional to ρ·W², so at 50% of design mass flow the DP is 25% of its span.
- True flow: 0.50 × 45,000 = 22,500 lb/h
- DP: 0.25 × 150 = 37.5 inH₂O → 25% of the 4–20 mA signal
- DCS, scaling mA linearly to lb/h with no square root: 0.25 × 45,000 = 11,250 lb/h
The indication is exactly half the true flow - the heat balance sees twice the indicated steam. The arithmetic matches the observation precisely.
Why commissioning missed it. At design flow the DP is at 100% and the linear scaling is also 100%, so indication and reality agree. As-built loop checks are almost always done at or near full scale. The error has the signature indication = f² where f is the true flow fraction, so it is invisible at design rate and grows as load falls: at 25% flow the indication shows 6.25% - a 75% error.
Confirm which end is wrong before changing anything. Two opposite faults are possible:
- Transmitter linear + DCS linear → indication = f², reads low at part load (this case).
- Transmitter square-root + DCS square-root → the two roots compound → reads high at part load.
Also note the meter's usable range: below roughly 25–30% of maximum DP, orifice uncertainty grows quickly, so a low-flow cutoff is required whatever the configuration.
2. Describe the verification you would perform before changing any configuration, and what you would check in the field versus in the DCS.
Read the worked answer
Record before you touch. Screenshot the DCS analog input and flow point configuration (transfer function, ranges, engineering units, damping) and read the transmitter's configuration (linear or square-root output, damping, sensor trim). A configuration change with no 'before' record is indefensible later.
DCS side - loop check with a mA source
Inject 4, 8, 12, 16, 20 mA at the DCS terminals and record the HMI value. If it reads 0, 25, 50, 75, 100% of 45,000 lb/h, then wiring and scaling are correct and the square-root extraction is genuinely absent. If it reads something else, the scaling itself is wrong - a second, separate fault.
Field side
- Zero check: with the 3-valve manifold equalized, the transmitter must read 4 mA / 0 inH₂O. Close the equalizer, open the high side, and confirm it responds.
- Impulse lines: check the high and low legs are correct, not swapped, with no condensate slugs or leaks, and the manifold valves seated properly.
- Transmitter transfer function: confirm it is linear as bench-calibrated. A transmitter left in square-root mode from a previous installation is a classic trap.
Process cross-check
Compare against the heat balance at three loads, or against a clamp-on ultrasonic meter or the condensate flow. Confirm the 2:1 relationship at half load rather than assuming it.
Datasheet check
Before trusting 45,000 lb/h, confirm from the orifice datasheet that 150 inH₂O really corresponds to 45,000 lb/h at the actual flowing pressure and temperature, and that the bore/β ratio matches the installed plate. If the plate was installed backwards or the bore differs, the fix above will produce a confident but wrong number.
3. Specify the correction, decide whether pressure/temperature compensation is needed here, and list the downstream consequences of the error.
Read the worked answer
The correction
- Apply the square root in exactly one place - preferably the DCS flow point, which also allows the low-flow cutoff. Update the loop drawing, I/O list, and instrument index to record which device extracts the root. Doing it in both is the classic double-root error described earlier.
- Re-range the HMI point to 0–45,000 lb/h with the square-root transfer and verify the totalizer uses the same corrected value.
- Set a low-flow cutoff at roughly 10–15% of span (or the datasheet's minimum usable DP) so that noise near zero DP is not amplified by the root and integrated by the totalizer.
Is compensation needed here?
Header pressure varies ±8 psi around 250 psig: 264.7 psia ±3.0%. For superheated steam, ρ ∝ P/T, and W ∝ √ρ, so a 3% density error produces a ±1.5% flow error - well inside the orifice's own uncertainty. Temperature is not controlled, though; if it swings 50 °F the density effect roughly doubles. Check the logged temperature spread before deciding.
Verdict: pressure/temperature compensation is optional for this duty, but required if the flow is used for custody transfer, energy allocation, or efficiency guarantees. In that case specify a flow computer and a documented compensation basis.
Downstream consequences
- The totalizer has under-registered for three months. Reconcile against boiler feedwater or condensate return, and correct any KPI, efficiency or billing figures derived from it.
- The reboiler temperature controller was tuned against a squared measurement, so its effective gain changed with load. Review and retune after the fix.
- Every alarm and interlock using this flow (low-flow trip, high-flow alarm) must be re-verified against the new scaling. Anything safety-related needs an MOC - a silently changed flow scale is exactly the kind of change that defeats a trip.